One Person, Two Accounts: The Smart Way to Test Genealogy Platforms

What if you could validate your family tree software without needing another relative. A simple dual-signup trick gives you reliable results even only one person signs up twice. The same principle that powers A/B testing in marketing departments works just as well for genealogy platforms like FamilySearch, Ancestry, and MyHeritage. The problem is practical: most genealogy tools assume a household of two when they test their matching algorithms.

But your research desk might hold exactly one living human with a login. You can sidestep that limitation entirely by creating two distinct accounts for the same individual. one with a maiden name, one with a married name, or two different email addresses. Why does this work. Because the core matching logic cares about records, not bodies.

When FamilySearch’s search index scans for a birth certificate from 1887 or MyHeritage compares a city directory entry against a census return, the algorithm sees strings of data. Two profiles under your control behave like two separate ancestors sharing DNA segments or record clusters. This approach isn’t faking anything; it’s stress-testing the machinery. If your dual accounts produce a confident match on a 1920 U.S.

Federal Census enumeration or an 1851 England census return from Findmypast, you know the platform handles ambiguity well. Failures reveal gaps in indexing coverage or overly strict fuzzy-matching thresholds.

Here’s what covers: step-by-step account setup that keeps both profiles legally compliant and technically distinct, then three benchmark tests to run across GEDmatch and FamilyTreeDNA’s free tiers. Before you open any registration page, grab a blank spreadsheet. one column for each account’s email domain (Gmail vs. Outlook, another for the birth date offset you’ll use between them. That small table becomes your reproducibility anchor when results come back unexpected.

Table of Contents

Why “Twin Tests” Fall Short

That reproducibility anchor solves half the problem. The other half is a common misconception: that validating record matching requires two distinct people with separate accounts. Family historians routinely assume meaningful results demand multiple testers, multiple trees, and weeks of coordination. The truth is simpler. One person, two accounts, and a carefully chosen birth date offset can produce results just as valid as any twin study.

This works because record-matching accuracy depends on the records themselves, not on who happens to be searching for them.

FamilySearch’s indexed collections contain billions of searchable records spanning census years 1790 through 1950 across the United States. When you submit a search query with identical parameters from two different accounts, the matching engine evaluates the same underlying index entries. The algorithm doesn’t know whether one human or two humans pressed the search button. What changes between accounts is your ability to control variables. A single researcher can vary one parameter at a time. birth year, surname spelling, county name.while holding everything else constant. Two people inevitably introduce unconscious differences in how they interpret instructions or structure their searches. The skeptical reader will ask about account fingerprinting. Platforms like AncestryDNA and MyHeritage do track user behavior, but their matching algorithms operate on record content, not browser metadata or typing cadence.

Your validation tests measure the quality of indexed records against your known genealogical facts. It works like peer review with yourself as both author and reviewer. You already know your great-grandfather’s exact birth date from his death certificate; what you’re testing is whether FamilySearch’s index agrees when you present that fact slightly differently each time. Start today by opening a second free account on FamilySearch using an alternate email address.

Set your first test: the 1900 United States Federal Census entry for your maternal grandfather’s family household in Cook County, Illinois.

Why Controlled Duplicates Reveal More Than That Cook County

A second profile holding the same family structure becomes your variable. Matching algorithms perform best when they face a known quantity. Submit the same maternal grandfather’s name, birth year, and birthplace on both records, then watch where each platform lands. FamilySearch’s search interface lets you toggle exact-match filters for surname, given name, and event year. The 1900 census enumeration districts in Chicago carry distinct page numbers, giving you a precise target to verify.

Your duplicate entry should introduce deliberate variations. Change the middle initial, swap a transcription error into the birth year, or list the wife’s maiden name instead of her married surname on one side. Matching engines treat these mutations differently. AncestryDNA’s record hints draw from its indexed collections across 20 billion entries, while FamilySearch’s free index relies on volunteer transcription quality that varies by county and decade.

The two profiles now function as a controlled experiment. When one entry produces a hint for that Cook County census page and the other does not, you have isolated how sensitive each algorithm is to data entry noise. Run this test against three record types: federal censuses from 1880-1940, state marriage indexes from Illinois, and city directories covering Chicago’s near-west side neighborhoods. Each collection presents different OCR challenges and indexing errors.

Record your findings in a simple spreadsheet with columns for platform, record type, variant used, and match result. A dozen trials across those collections will outperform a hundred vague searches on one database alone. Do not settle for a single pass. Reverse the experiment by making your primary profile the altered one next week; inconsistent results between runs signal where an algorithm’s confidence thresholds wobble under pressure.

Your next step: Pick one Cook County ancestor and create that second profile today. start with FamilySearch’s free account to keep costs at zero while you build your baseline data.

Coverage vs Consistency: Two Different Tests

The reverse pass exposes something deeper than a glitch. It separates data coverage from algorithmic consistency. Coverage asks whether a record exists in the database at all; consistency asks whether two identical searches return the same result. Most researchers conflate them, and that confusion invalidates their conclusions. Coverage is the easy test. Search FamilySearch’s catalog for an 1880 Ohio marriage index, then repeat the query on Ancestry’s collection set.

The differences you find are archival gaps: missing counties, unindexed years, illegible originals that never made it to microfilm. Take one clean record. Census entry for your great-grandmother. and search it ten times across MyHeritage using slightly varied name spellings: “Katherine,” “Catharine,” “Katy.” The database either surfaces the same household each time or it doesn’t.

A platform can ace one and fail the other spectacularly. Findmypast may hold every parish register for Lancashire (superb coverage) while its fuzzy-matching algorithm ranks a transcription error above your exact surname hit (inconsistent ranking). Neither failure cancels out the other. FamilySearch’s free index suffers from volunteer transcription quirks. “Smith” becomes “Smyth” in one batch and “Smithe” in another. yet its search returns both variants reliably once you account for soundex coding.

MyHeritage applies stronger normalization but may occasionally merge distinct individuals with similar birth years. Your twin-less methodology works because each sign-up gives you a fresh algorithmic slate. The second account skips your cached session history, forcing a cold-start query against the same underlying index. Run both accounts against identical search strings on Findmypast; if ranking order shifts between logins, you’ve found a session-dependent scoring artifact rather than a fixed record truth.

Document your results in a simple spreadsheet with columns for database name, record type, query string, and result order per account. Three rounds of testing across FamilySearch and Ancestry will reveal which platforms treat identical queries as deterministic functions of their index. and which inject variability that undermines genealogical reproducibility.

Setting Up Two Accounts Without Contaminating Results

Duplicate profiles only compromise a test when the platform can link them. FamilySearch ties profiles to a single login email, but AncestryDNA and MyHeritage permit multiple trees per person. The trick is making each profile distinct enough to pass automated checks while keeping the underlying DNA data identical. Start with email: create one address through Gmail’s plus-addressing system (yourname+test1@gmail.com) and another through a free ProtonMail account. Both receive verification links without exposing your primary inbox.

Name formatting provides your second layer of separation. Use your full legal name on Profile A. and a nickname variant on Profile B. Whitfield.” Census records from 1850 through 1940 show exactly this pattern; ancestors routinely swapped between formal and familiar names across different documents. The autosomal DNA test itself needs no modification.

A single spit kit from AncestryDNA costs \(99 and yields one raw data file that you upload to both profiles via GEDmatch’s free tier or FamilyTreeDNA’s transfer feature (currently \)19 for autosomal matching). The DNA never changes; only the metadata around it does. Avoid common pitfalls that trigger platform reviews. Do not use the same phone number for both registrations. Google Voice offers a second number for free in most U.S.states as of 2026, which sidesteps SMS verification conflicts entirely. Similarly, vary your birth date by one day or one month between profiles; algorithms flag identical demographic strings faster than they flag the underlying genotype. Record your setup in a spreadsheet before testing begins. Column headers should include: profile name, email address, associated phone number, birth date variant, and the date you uploaded each raw DNA file (typically available within 3. 5 business days after AncestryDNA receives your kit.

This audit trail prevents you from confusing your own two identities mid-project. Your results remain statistically valid because you control both variables: identical genetic input with deliberately differentiated registration data. Genetic genealogist Blaine Bettinger’s Genetic Genealogy in Practice workbook covers controlled duplicate testing for surname matching. Start today by ordering one AncestryDNA kit and setting up that ProtonMail address before it arrives. You will have isolated the account-creation variable within an hour of receiving your raw data download link.

Keep the First Inputs Identical

The raw data download arrives, and your second account stands ready. Now comes the discipline that separates a real experiment from a self-deception exercise. Every field you populate on that second profile must mirror the first exactly. same birth date, same place name, same spelling of your great-grandmother’s maiden name. FamilySearch’s record hints and AncestryDNA’s ethnicity estimates both key off these foundational identifiers, and a single transcription difference muddies which system generated the variance.

Build a reference sheet before you type anything. List each input field as it appeared on account one. given names, surnames, locations, dates. and transcribe them character-for-character onto account two. This takes roughly 15 minutes and eliminates input drift as an explanatory variable.

One deliberate exception exists: change your email address and display name only. ProtonMail gives you that clean identifier without touching genealogical data fields. The systems under test respond to biographical strings and relationship links, not your login credentials. Blaine Bettinger’s workbook reinforces that duplicate testing requires ruling out input differences before attributing results to algorithmic behavior. His controlled trials across FamilyTreeDNA and GEDmatch showed identical profiles returning divergent surname match rankings within weeks. That divergence is your signal.

When two identical profiles produce different matches or hint orders on MyHeritage versus AncestryDNA, you have documented proof of platform logic at work rather than sloppy preparation. Confirm both accounts accept the same DNA file format before uploading. GEDmatch accepts raw data from all major vendors, but if you are comparing two kits within one platform like AncestryDNA, verify both are autosomal tests processed through the same chip generation.older v1 chips versus newer v2 chips can introduce noise unrelated to your research question. Tomorrow morning, upload that second raw file with your reference sheet beside the keyboard. The comparison window opens the moment both kits show “results ready”. and now you can trust every difference it reveals.

Duplicate Smart, Not Just Twice

That trust extends beyond the raw file itself. When you register your second profile, the settings you pick can quietly corrupt your comparison. Or keep it pristine. Your notification choices deserve the most attention. Log into FamilySearch and set both profiles to “email me only for critical alerts” before uploading anything; default digests will flood your inbox with matching hints that skew your perception of what’s genuinely new versus merely re-indexed. Record preferences matter less than you think.

Whether you’ve toggled “show me Irish Catholic parish records first” or left the discovery defaults untouched, those choices shape search suggestions but not the underlying DNA match calculations. The algorithm reads your genome identically either way. What actually contaminates results is privacy sharing. On both FamilySearch and MyHeritage, check that “allow others to view my tree” remains off until both kits finish processing. Accidental cousin connections create false-positive matches that look like new discoveries. Double-check location data too.

If Profile A lists Chicago and Profile B lists Boston, some platforms weight geographic proximity in their match rankings, manufacturing a difference that isn’t genetic at all.

Here’s the pragmatic workflow: 1. Create Profile B with a fresh email address (Gmail’s + trick works: yourname+test@gmail.com. Disable all email notifications immediately after registration. Set both profiles to private visibility under Settings > Privacy. Upload Kit B using the same physical sample tube, just different login credentials. Wait for both results pages to show “ready” before opening any comparison view.

The test card from AncestryDNA ships with a unique activation code per kit; order two kits under two names and note which code maps to which profile on paper before scanning either barcode into FamilyTreeDNA or GEDmatch. One last setting separates clean research from muddled noise: chromosome browser export preferences. Keep GEDmatch’s default “raw segments only” rather than enabling enhanced visualization tools. they add interpretation layers no control subject needs.

Write down your chosen privacy configuration on a sticky note and tape it beside your monitor this afternoon. Thirty seconds of settings hygiene now saves you three hours of untangling phantom matches later.

Why This Test Works

The dual-account method works because genealogy platforms match records, not identities. You are not gaming the system; you are auditing it. Two profiles under your control behave like two separate ancestors sharing record clusters, and the matching engine evaluates the same underlying index entries regardless of whether one human or two pressed the search button. What changes between accounts is your ability to control variables: birth year, surname spelling, county name.

A single researcher can vary one parameter at a time while holding everything else constant, producing results just as valid as any twin study. The discipline that makes this work is simple: keep every biographical field identical across both profiles—same birth date, same place name, same spelling of your great-grandmother’s maiden name—while changing only email, display name, and privacy settings.

Record your setup in a spreadsheet before testing begins, with columns for profile name, email address, birth date variant, and upload date. That audit trail prevents you from confusing your own two identities mid-project. When two identical profiles produce different matches or hint orders on MyHeritage versus AncestryDNA, you have documented proof of platform logic at work rather than sloppy preparation.

Run three concrete trials to see this in action. First, search the 1900 U.S. Federal Census for your maternal grandfather’s household in Cook County, Illinois, from both profiles using identical parameters; if FamilySearch returns the same enumeration district page for both, the index is consistent.

Second, search an 1880 Ohio marriage index with a deliberate surname misspelling on one profile only; if the corrected spelling surfaces the record but the misspelled version does not, you have isolated the algorithm’s fuzzy-matching threshold. Third, upload the same raw DNA file to GEDmatch under both profiles and compare surname match rankings; if the top ten matches shift order between logins, you have found session-dependent scoring variability.

A platform can hold every parish register for Lancashire yet rank a transcription error above your exact surname hit. Neither failure cancels out the other. Your twin-less methodology works because each sign-up gives you a fresh algorithmic slate, skipping cached session history and forcing a cold-start query against the same underlying index. If two profiles you control connect across a census return and a baptism record, that match is genuine evidence the indexing pipeline holds up.

So grab a spare email address and create that second profile today. Run one comparison on FamilySearch or MyHeritage with an ancestor who has a common name. The result will tell you more about your software’s limits than a dozen hours of manual searching ever could. Your research deserves that clarity, and now you know exactly how to test for it.